In this demo for Qualcomm AI PCs (Snapdragon X Elite and X2 Elite), I’m walking through two simple but powerful on-device agent workflows in Model HQ—the same demo Qualcomm showcased at CES Las Vegas (January 2026). Everything is built in our no-code, drag-and-drop Visual Builder, and runs locally on device for a private, secure workflow. ✅ Demo #1: Document Agent (RAG + Chat + Calculation) We start with a music license agreement and build a lightweight RAG workflow to answer: Who are the parties to the agreement? What is the royalty rate? Then we use a Transformer node to bundle the answers into Agent State, and pass it into a final chat step to calculate: What is the royalty if the artist sells 100 copies at $5 each? To showcase the Qualcomm NPU, we run the calculation using a Foundry model (Qwen 2.5 7B Instruct). 🖼️ Demo #2: Vision + Classifiers + Storytelling Agent Next, we build a fun vision workflow using an image input: Vision model describes the image Sentiment classifier detects positivity/negativity Emotion classifier detects emotion Transformer bundles outputs → chatbot writes a short story We run this workflow with a Foundry model (Phi-3) and test it with: A dog running through a field 🐶 A sports car image 🚗 🔥 Why this matters In a single workflow, we’re chaining 4 models together (vision + sentiment + emotion + chat) and showing how CPU + NPU can work together seamlessly on Qualcomm hardware. 📌 Model HQ: Private • Local • No-Code • Built for Agents + Small Models #Qualcomm #SnapdragonXElite #OnDeviceAI #NPU #Agents #RAG #VisionAI #ModelHQ #FoundryModels #CES2026 #LLMWare #OnDeviceAI #PrivateAI #EdgeAI #AIDemo